Early Decision in a Censored Wilcoxon Two-Sample Test for Accumulating Survival Data

For two-sample testing problems, in which observations are collected in order of size, Ailing [1] showed that it is possible to determine the outcome of an uncensored Wilcoxon test well before data collection is complete. This article studies a similar problem for a censoring scheme terminating the experiment when a specified percentage of events has occurred in either group. The asymptotic distributions (null and nonnull) of the test statistic (of Wilcoxon type) and of the proportion of events after which early stopping occurs are reported, and some Monte Carlo work is compared with Alling's results. Other specifications of censoring and summary statistic and some applications are discussed.